Optimization of a renewable energy supply system on a remote area:

Berlenga Island case study

L. Amaral1*, N. Martins2 and J. Gouveia3
1*
Department of Economics, Management and Industrial Engineering; Research Unit on Governance,
Competitiveness and Public Politics (GOOVCOP) University of Aveiro; Campus Universitário de Santiago,
3810-193 Aveiro, Portugal
2
Department of Mechanical Engineering; Centre for Mechanical Technology and Automation (TEMA);
University of Aveiro; Campus Universitário de Santiago, 3810-193 Aveiro, Portugal
3
Department of Economics, Management and Industrial Engineering; Research Unit on Governance,
Competitiveness and Public Politics (GOOVCOP); University of Aveiro; Campus Universitário de
Santiago, 3810-193 Aveiro, Portugal
*Correponding Author. Tel: +351 234 370 830, Fax: +351 234 370 953, E-mail: lpgamaral@ua.pt

Abstract: HOMER software was used to technical and economically assess two renewable energy supply (RES)
system configurations – PV-only (80 kW) and PV (34 kW)/WT (36 kW), both with a three day storage capacity
and requiring 11kW of electric power – proposed by a partnership project responsible for the implementation of
sustainable measures on a Portuguese small island. HOMER calculation showed insufficient storage capacity for
both RES system proposed, so extra storage capacity should be added. Economically, life cycle cost (NPC) of
the cheaper configuration (PV/WT) resulting from HOMER calculation was significantly lower (20%) than the
one advanced by the project. On a second stage, HOMER was used to compute an optimal RES system
configuration to attend water desalination and street lighting electric additional loads. The optimal configuration
– PV (25 kW)/WT (18kW) – costs 18% less than the equivalent PV/WT system proposed by the project when
the same additional load is considered. Sensitivity analysis on the electric load showed the cost difference
between project’s and HOMER’s proposals fading as the load increased. Variation on wind speed average
demonstrated the significance of data accuracy: using NASA’s average wind speed data the NPC increased on
15% compared to using wind speed values revealed on a monitoring campaign on the island.

Keywords: Remote off-grid energy systems, Optimization software, Sensitivity analysis

1. Introduction
Decentralized energy generation systems have become a recent trend on the development of
energy systems. Concerns related to energy security and climate change have been fostering
the implementation of projects that allow the production of power and heat closer to the point
of use [1], as the current and dominant approach of centralized energy production, based on
fossil fuels, lead to inequities, external debts and significant environmental degradation [2].
Sustainability of energy systems is based on the energy hierarchy principles: top priority is
energy conservation, next the adoption of renewable resources for energy production, and last
the use of fossil resources [3].

Planning energy systems represents a major issue on the development process of our society.
Available computational energy models can support energy planners to decide the best
configuration for an energy supply system, as they allow simulating different solutions and
working conditions, and checking their technical and economical feasibility in an early stage
of the decision process. Optimization models, namely linear programming mathematical
models, are usually used to solve cost minimization problems subject to specific
technological, political and demand satisfaction constraints given by energy models [4]. This
is the case of HOMER®, a software developed by the US National Renewable Energy
Laboratory to address the need for a hybrid system design tool accurate enough to predict
energy system performance. It has been used on several situations all over the world: a
feasibility study for the implementation of a zero home energy in a Canada’s city [5]; study of

3484
wind penetration into an existing diesel plant of an Saudi Arabia village [6]; analysis of the
technical and financial viability of grid-only, renewable energy supply (RES)-only and
grid/RES hybrid power supply configurations for a large-scale grid-connected Australian
hotel [7].

2. Methodology
2.1. HOMER Software
HOMER is primarily an optimisation software package which simulates different RES system
configurations and scales them on the basis of net present cost (NPC), which is the total cost
of installing and operating the system over its lifetime. Depending on the input data and
constrains imposed by the user, HOMER firstly assesses the technical feasibility of the RES
system (i.e. whether the system can adequately serve the electrical and thermal loads and any
other constraints imposed by the user), and then estimates the system’s NPC [7]. Besides the
electric load to attend, the user has to specify the “search space”, i.e., the sizes and/or
quantities of the different components of the RES system (wind generators (WT),
photovoltaic array (PV), batteries, inverters, electrolyser, generator…) that will be used to
calculate the optimal system design. It also performs sensitivity analysis to evaluate the
impact of a change in one or more of the input parameters.

HOMER was used on this paper with three purposes: first, to assess the technical and
economical performance of two predetermined RES system configurations; second, to
optimise a RES system based on wind and solar resources on the Berlenga Island; and third,
to assess the impact of the variation on electric load and the average wind speed has on the
optimal RES system configuration.

2.2. Case Study: Berlenga Island

The Berlengas Archipelago is located 6 miles away from the Carvoeiro cape on Western
Portugal, and has approximately 100 ha. Its island is called Berlenga Island. There is no
resident population on this small group of islands, which contributed for the preservation of
singular species of flora and fauna. Despite the absence of resident population, there is some
human activity on Berlenga Island all year long: lighthouse workers are present on the island
24H/7day during all year. They work on rotation teams and spend several days in a row all
year long; some Peniche’s municipality workers spend some periods of time on the island
form March to November; from May to October nearly 30 fisherman and restaurant workers
stay full time on the island. Besides these “permanent” residents, there is a legal limitation of
350 islands visitors [8].

Before 2007, when a partnership program called “Berlenga – Sustainability Lab” (from now
on called Berlenga Project) started, electricity generation was based on diesel generators (130
kW), producing 30 MWh/year and consuming nearly 15000 L/year (roughly 40 ton CO 2
emissions/year) [9]. This system had several drawbacks: high O&M diesel costs due to the
aggressive environment on the island; limited energy supply schedule; island development
compromised due to electrical limitations (water treatment systems) [9-10]. As so, Berlenga
Project intended to develop a zero CO 2 emission electric system to supply the Berlenga
Island. Two possible configurations were proposed: Configuration A – PV/WT system;
Configuration B – PV-only system. Accordingly to Berlenga Project, Configuration A is less
expensive than Configuration B (600 k€ and 750 k€, respectively), however it has an higher
environmental impact (mainly because of WT’s visual impact and sound pollution) [10].

3485
2.2.1. Berlenga Project Electric Load
In an early stage of Berlenga Project, island’s electric monthly load profile was monitored –
Fig. 1. Island’s electric load profile shows irregular electricity consumption due to the “tourist
invasion” during summer months. This domestic electric load profile refers to electricity use
on households only. From December to February the only residents on the island are the
lighthouse workers. The lighthouse already had a PV array installed and that justifies the
absence of electric load on those months. To perform a HOMER simulation, a monthly load
profile is not accurate enough. It is required an hourly electric load profile. For that purpose it
was used a typical household hourly load profile – Fig. 2 – in order to calculate an hourly load
for an average day of each month of the year. The maximum electric power considered was
11 kW and the electric load is subject to 5% standard deviation on daily averages and 10%
deviation between the difference of hourly data and the average daily profile.

6000 210

Daily Average (kWh)

5000 175
Average (kWh)

4000 140 Average

Monthly

monthly
3000 105

2000 70 Average
daily
1000 35

0 0
Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec

Figure 1. Monthly and daily average domestic electric load profile – adapted from [9].

10,0

8,0

6,0
% Load

4,0
(kW)

2,0

0,0
00 - 01
01 - 02
02 - 03
03 - 04
04 - 05
05 - 06
06 - 07
07 - 08
08 - 09
09 - 10
10 - 11
11 - 12
12 - 13
13 - 14
14 - 15
15 - 16
16 - 17
17 - 18
18 - 19
19 - 20
20 - 21
21 - 22
22 - 23
23 - 00

Hour
Figure 2. Hourly load percentage distribution during a typical day – adapted from [11].

2.2.2. Additional Electric Loads

Installation of street lighting on the island was assessed on this study. The street lighting
system should include 10 street lamps with 125W each working on average 10h/day (3650
hours/year). During winter (Oct-Mar) street lightning works 12 hours/day (from 19:00 to
07:00) and on summer period (Apr-Sep) only 8 hour/day (from 22:00 to 06:00).

The island doesn’t have fresh water reserve aquifers. Fresh water is supplied to the island
through an 8 m3 container, by the ship that takes the visitors to the island [12]. According to
the last report on this matter [12], fresh water consumption on the island during high season
(July and August) was 3 m3/day and 2 m3/day on 2007 and 2008 respectively. The fresh water
load and electric load required to produce it using a reverse osmosis equipment, are shown on
Fig. 3. This electric load was considered as a deferrable load, i.e. electrical load that must be

3486
met within some period of time, but the exact timing is not important.

100 625

Electricity requirement
80 500
Water Load (m3)

Fresh Water
60 375 Load

(kWh)
40 250 Required
Electricity
20 125

0 0
Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec

Figure 3. Fresh water consumption and required electricity to produce it.

2.3. Renewable Resources – Wind and Sun

Two sources of information were available on wind average speed. Through NASA’s website
on data information on surface meteorology and solar energy [13] it’s possible to get a
monthly profile according to the geographic coordinates of a certain location. To the location
of Berlenga Island the database indicate a baseline annual average of 5,16 m/s. In addition to
this information, wind potential on the island was evaluated through a monitoring campaign
from December 2006 to September 2007 [14]. The campaign’s reported an average wind
speed of 6,74 m/s, significantly higher than NASA’s baseline annual average. As so, it was
used 6,74 m/s as a scaled annual average value on the simulation. The scaled data retains the
shape and statistical characteristics of the baseline data, but may differ in magnitude.

Solar data information provided by HOMER database was used, according to the geographic
coordinates of the island – 4,092 kWh/m2/day and 0,518 Clearness Index.

2.4. Equipments
To perform HOMER simulations, RES system equipments shown on Table 1 were
considered.

Table 1. Capital and operation and maintenance costs of a RES system’s equipment.
Equipment Capital Cost O&M Cost
6 kW Wind Turbine 29445 €/unit [15] 600 €/year [16]
15 kW Wind Turbine 76700 €/unit[15] 850 €/year [16]
Photovoltaic Array 5500 €/kW [17] 10 €/kW [18]
Battery 6CS – 7,6 kWh 850 €/unit [19] 10 €/year [19]
Battery 4KS – 6,94 kWh 770 €/unit [19] 10 €/year [19]
Inverter/Converter 550 €/kW [19] -

3. HOMER simulation results

Solar and wind data, electric loads and equipments required to build a RES system were
described on the previous sections. The project lifetime is 25 years.

3.1. Berlenga Project Configurations

The Berlenga Project originally proposed two alternative configurations for the required RES
system: Configuration A – PV (34 kW)/Wind Turbine (36 kW); Configuration B – PV (80
kW). In order to meet Berlenga Project constrains, both configurations should attend the

3487
required domestic electric load and include batteries with capacity to bear three days of
average consumption, i.e., 230 kWh [9-10]. After running the simulation with this storage
capacity, the only feasible configuration resulted on a PV (80 kW)/WT (36 kW) system,
mainly due to high electric demand on summer days. As so, it was necessary to extend the
storage capacity to enable the simulation of both Configurations A and B – Table 2.

3.2. Additional electric load scenario

Configuration A and B were simulated to attend the domestic electric load monitored on the
scope of the project as shown on Fig. 1. It was simulated a RES system configuration to
attend additional street lighting and water desalination electric loads – Current Load Scenario
– resulting on Configurations C and D, as shown on Table 2. Adding these two additional
loads resulted in higher Initial Capital Cost (IC) and NPC, basically due to the requirement of
extra storage capacity and respective O&M costs, however the cost of energy (COE) dropped.
HOMER defines COE as the average cost per kWh of useful electrical energy produced by
the system. The lower COE shows more efficient use of the electricity produced, as the same
amount of electricity is produced and more energy is effectively used.

Table 2. RES systems configurations proposed by Berlenga Project.

RES PV WT 6kW 6CS 4KS Conv IC O&M NPC COE
Config (kW) (unit) (unit) (unit) (kW) (k€) ($/yr) (k€) ($/kWh)
3 day
80 6 30 21 642 8789 764 1,415
storage
A 34 6 35 19 390 5744 506 0,937
B 80 65 21 495 4168 579 1,073
C 34 6 40 19 398 6101 520 0,757
D 80 80 21 508 4874 606 0,882

3.3. Optimal configuration by HOMER

As an alternative solution to the originally proposed configurations, above referred as A and
B, HOMER was used to shape an optimal RES system based on PV panels, wind turbines
and a battery bank to store electricity in order to attend the same domestic electric load shown
on Fig 1. It was included a 15 kW wind turbine on this simulation in addition to the 6kW
wind turbine. The optimization results are shown on Table 3. Three different configurations
result to be possible for the implementation of a RES system on the Berlenga Island:
Configuration 1 – PV (25kW) + WT (3x6kW); Configuration 2 – PV (65 kW); Configuration
3 – WT (9x6kW). Configuration 1 represents the HOMER optimal configuration – lowest
NPC. The RES system based on PV-only (Configuration 2) is significantly more expensive
than mixing PV and WT. A third option (Configuration 3) is available using wind turbines
only. This one has a lower IC than configuration 2, but the higher O&M costs results on a
higher NPC.

Table 3. HOMER optimal configuration when attending the domestic, water desalination and street
lighting loads.
RES PV WT 6kW WT 15kW 4KS Conv IC O&M NPC COE
Config (kW) (unit) (unit) (unit) (kW) (k€) ($/yr) (k€) ($/kWh)
1 25 3 85 22 298 6374 426 0,620
2 65 95 21 438 5430 548 0,796
3 9 100 19 350 10387 559 0,813

3488
Comparing Berlenga Project’s original configurations C (PV/WT) and D (PV-only) with
configurations 1 (PV/WT) and 2 (PV-only) resulting from HOMER optimization, it is clear
that the last ones present better financial indicators, especially the PV/WT configuration.

3.4. Sensitivity Analysis by HOMER

3.4.1. Electric Load
The effect of higher electricity requirement on HOMER optimal configuration was assessed
through a sensitivity analysis, by creating two additional electric load scenarios: 10% Increase
Scenario – 10 % increase on domestic and water desalination electric load; 20% Increase
Scenario – 20 % increase on domestic and water desalination electric load. Street lighting
remained unchanged on both scenarios. A PV/WT system resulted to be the optimal
configuration for both scenarios. The NPC increases linearly with the electric load – Fig. 3.
Comparing configuration C (34 kW PV/36 kW WT) with HOMER optimal configuration for
each electric load scenario, it can be stated that the greater the electric load the closer is the
economic performance of both configurations even though the optimal configuration
proposed by HOMER is always cheaper – Fig. 4.

With 10% load increase scenario the three possible configurations for the RES system showed
on Table 3 are feasible. As the electric load increase 20%, PV-only solution becomes
unfeasible.

600
500 Current Load
400
k€ 10% load
300 increase
200 20% load
increase
100
00
Initial capital Total NPC

Figure 4. Effect of increasing electric load on RES system costs.

600
500
Configuration C
400

k€ 300
200
HOMER Optimal
100 Config
0
current load 10% load increase 20% load increase

Figure 5. NPC comparison between Configuration C and Configuration 1.

3.4.2. Wind speed

As occurred with the electric load sensitivity analysis, a PV/WT system resulted to be the
HOMER optimal configuration when assessing wind speed influence. The economic impact
of the average wind speed on the optimal RES system can be assessed on Fig. 5. Using
NASA’s database value for annual average wind speed, NPC increased around 11% on all
three electric load scenarios.

3489
 600

500
NASA Wind
400
Speed
k€ 300

200 Monitored
Wind Speed
100

0
current load 10% load increase 20% load increase

Figure 6. Effect of the average wind speed on system’s NPC for each of the electric load increase.

For the current load scenario, the feasibility of the three possible configuration systems is not
affected despite the WT-only system results to be almost 80% more expensive than the
PV/WT optimal configuration. As the load increase 10% the WT configuration becomes no
longer feasible and with 20% electric load increase PV/WT is the only feasible configuration.

4. Conclusions
One of the assumptions made by Berlenga Project’s authors was to use a three day storage
capacity (230 kWh). HOMER’s system simulation showed that this storage capacity not
enough for the proposed RES system configurations (A and B) mainly because of the high
demand on summer months.

Information on system’s cost issued by the Berlenga Project is not completely obvious. A cost
of 600 k€ for configuration A (PV/WT) and 750 k€ for configuration B (PV-only system) was
assumed, but it is not clear if those are IC or NPC [9]. Assuming those values as NPC and that
the aim is to attend the domestic electric load monitored under the Berlenga Project, the costs
resulting from HOMER simulation for both system configurations are lower than those
proposed by the Berlenga Project: 506 k€ for configuration A and 579 k€ for configuration B.
When attending additional street lighting and the water desalination loads (configurations C
and D) the costs rose, due to higher storage capacity needed, but they still were far from the
costs advanced by the Berlenga Project. This cost gap can be explained, at some extent, by the
fact that the costs advanced by the Berlenga Project were from 2007, three years ago. When
performing the “free” optimization to attend domestic, water desalination and street lighting
loads (Current Load Scenario), the optimal configuration (configuration 1) includes 25 kW
photovoltaic panels and three 6 kW wind turbines. Both PV/WT and PV-only optimized
configurations resulted to be cheaper than the ones proposed by the Berlenga Project.

Two sensitivity analyses were performed to assess the impact of electric load deviation and
average wind speed on the RES system cost and configuration attending domestic, street
lighting and water desalination loads. As the electric load rises, the closer is the economic
performance of both Berlenga Project (configuration C) and HOMER optimal configurations,
even though the latter is always cheaper. This means that the configuration proposed by the
Berlenga Project was oversized for the present electric load requirement. This may be an
assumed choice, having in mind the future load growth, when new electric loads were
included in the grid, the only equipments to add on the RES system should be batteries to
store electricity. Despite using NASA’s data to profile the average monthly wind speed, a
scaled value for the wind speed based on the monitoring campaign made on the Berlenga
Island was used to compute the simulation, for a better representation of local conditions.
Higher average wind speed means more available wind resource and less costs to generate the
same amount of electricity. The higher wind speed value results on a 10% lower NPC for the
optimal PV/WT configuration. This is true both for current load and for load increase

3490
scenarios. This analysis stresses the importance of use valid reliable data, namely renewable
resources availability data, when performing a technical and/or economical assessment of a
RES system.
References
[1] Wolfe, P., The implications of an increasingly decentralised energy system. Energy
Policy, 2008. 36(12): p. 4509-4513.
[2] Hiremath, R.B., S. Shikha, and N.H. Ravindranath, Decentralized energy planning;
modeling and application - a review. Renewable & Sustainable Energy Reviews, 2007.
11(5): p. 729-752.
[3] Association, R.E., et al., Sustainable Energy Manifesto. 2006.
[4] Antunes, C.H., A.G. Martins, and I.S. Brito, A multiple objective mixed integer linear
programming model for power generation expansion planning. Energy, 2004. 29(4): p.
613-627.
[5] Iqbal, M.T., A feasibility study of a zero energy home in Newfoundland. Renewable
Energy, 2004. 29(2): p. 277-289.
[6] Rehman, S., et al., Feasibility study of hybrid retrofits to an isolated off-grid diesel power
plant. Renewable and Sustainable Energy Reviews, 2007. 11(4): p. 635-653.
[7] Dalton, G.J., D.A. Lockington, and T.E. Baldock, Feasibility analysis of stand-alone
renewable energy supply options for a large hotel. Renewable Energy, 2008. 33(7): p.
1475-1490.
[8] Portaria nº 270/90 de 10 de Abril, M.d.A.e.R. Naturais, Editor. 1990.
[9] Maciel, J., Berlenga Sustainability Laboratory - A free-Carbon Energetic Solution. 2007.
[10] Estanislau, S. Berlenga – Sustainability Lab/Plans. in 2007 C3P & NASA Technical
Workshop Partnership for Energy and Environmental Stewardship. 2007. Peniche.
[11] Fritz, W. and D. Kallis, Domestic Load profile measurtements and analysis across a
disparate consumer base, in 18th International conference on the "Domestic Use of
Energy". 2010.
[12] Correia, C.S., Projecto “Berlenga – Laboratório de Sustentabilidade” Soluções de
Abastecimento de Água e Saneamento, in Departamento de Ciências e Engenharia do
Ambiente. 2008,: Lisboa.
[13] NASA. Surface meteorology and Solar Energy. A renewable energy resource web site
(release 6.0) sponsored by NASA's Earth Science Enterprise Program Available from:
http://eosweb.larc.nasa.gov/sse/.
[14] Simões, T., P. Costa, and A. Estanqueiro. Wind resource assessment in Berlengas Island.
in European Wind Conference. 2007. Berlin, Germany.
[15] Proven Energy website. 2008; Available from:
http://www.provenenergy.co.uk/our_products.php.
[16] Canada Wind Energy Association Website. 2008; Available from:
http://www.smallwindenergy.ca/en/Overview/Costs/CostComparison.html.
[17] Suneffect website. 2009; Available from: http://www.suneffects.net/index_m.html.
[18] Partnership, P.R. Public Renewables Partnership website. Available from:
http://www.repartners.org/solar/pvcost.htm.
[19] SolarBuzz. SolarBuzz website. 2010; Available from: http://www.solarbuzz.com.

3491